Conservation Genetics

, Volume 9, Issue 3, pp 653–666

Cryptic species diversity in a widespread bumble bee complex revealed using mitochondrial DNA RFLPs

Authors

    • School of Biological SciencesQueen’s University Belfast
    • Crops Research Centre
  • Úna Fitzpatrick
    • Department of Biology, School of Natural SciencesUniversity of Dublin Trinity College
  • Mark J. F. Brown
    • Department of Biology, School of Natural SciencesUniversity of Dublin Trinity College
  • Robert J. Paxton
    • School of Biological SciencesQueen’s University Belfast
Research Article

DOI: 10.1007/s10592-007-9394-z

Cite this article as:
Murray, T.E., Fitzpatrick, Ú., Brown, M.J.F. et al. Conserv Genet (2008) 9: 653. doi:10.1007/s10592-007-9394-z

Abstract

Cryptic species diversity is thought to be common within the class Insecta, posing problems for basic ecological and population genetic studies and conservation management. Within the temperate bumble bee (Bombus spp.) fauna, members of the subgenus Bombus sensu stricto are amongst the most abundant and widespread. However, their species diversity is controversial due to the extreme difficulty or inability morphologically to identify the majority of individuals to species. Our character-based phylogenetic analyses of partial CO1 (700 bp) from 39 individuals spread across their sympatric European ranges provided unequivocal support for five taxa (3–22 diagnostic DNA base pair sites per species). Inclusion of 20 Irish specimens to the dataset revealed ≥2.3% sequence divergence between taxa and ≤1.3% within taxa. We developed a PCR-RFLP based method for unequivocally distinguishing amongst the four cryptic European taxa of this subgenus, B. cryptarum, B. lucorum, B. magnus and B. terrestris, and used it to analyse 391 females of the former three species collected across Ireland, all of which could be unambiguously assigned to species. Bombus lucorum was the most widely distributed and abundant of the cryptarum–lucorum–magnus species complex, comprising 56% of individuals, though it was significantly less abundant at higher altitudes (>200 m) whilst B. cryptarum was relatively more abundant at higher altitudes. Bombus magnus was rarely encountered at urban sites. Both B. lucorum and B. terrestris are nowadays reared commercially for pollination and transported globally. Our RFLP approach to identify native fauna can underpin ecological studies of these important cryptic species as well as the impact of commercial bumble bees on them.

Keywords

BombusCytochrome c oxidase subunit 1CO1PCR-RFLPDNA barcode

Introduction

The taxonomic status of a species must be accurately described before its conservation status can be established. Yet even in well-known taxa the presence of morphologically unrecognised (or cryptic) species indicates that there are many more species present than currently estimated (Frankham et al. 2002; Blaxter 2004; Bickford et al. 2007). Failure to account for cryptic species in widely distributed organisms has alarming implications for their management; the causal link between environmental variation and changes in species distribution may be lost and significant units for conservation may be erroneously delineated (Riddle et al. 2000).

Cryptic species are a particular problem in the most species rich animal group, the insects (e.g. Wells and Henry 1998). DNA barcodes, short DNA sequences, have recently been advocated as one means of identification which, in the absence of other data, allow species to be designated on the basis of genetic distance between species (Hebert et al. 2003a, b; Savolainen et al. 2005). However, lack of consensus on a standardised percent sequence divergence among species, in combination with a relatively high frequency of overlap in intraspecific diversity versus interspecific divergence in some groups, has undermined the utility of the approach in species discovery and species delimitation (Cognato 2006; DeSalle et al. 2005; Rubinoff 2006). As advocated by DeSalle et al. (2005)and DeSalle (2006), character-based systems, in which DNA sequence identity is used to generate discrete diagnostic characters, overcome the obstacles of distance-based barcodes and are more appropriate for species identification. The DNA character-diagnostic approach compliments the way taxonomists collate morphological or behavioural metrics when describing species, facilitating independent verification of taxonomic hypotheses; in cases where other taxonomic data are lacking or sequence information conflicts with current taxonomy, the DNA character-diagnostic approach can generate new hypotheses about the taxonomy of the group (DeSalle et al. 2005; Savolainen et al. 2005; Bickford et al. 2007).

Bumble bees (Hymenoptera: Apidae, genus Bombus) are of great ecological and economic importance in temperate terrestrial ecosystems as major pollinators yet they are a currently declining group (Williams 1986, 2005; Kearns et al. 1998; Goulson 2003; Goulson et al. 2005, 2006; Carvell et al. 2006; Fitzpatrick et al. 2007). For example, six of Ireland’s 20 Bombus species are on the recently established IUCN regional red data list of Irish bees (Fitzpatrick et al. 2006). In the field, species are identified primarily on small morphological differences, especially in coat colour and pilosity (Alford 1975; Prŷs-Jones and Corbet 1991; von Hagen 2003; Edwards and Jenner 2005; Benton 2006; Feltwell 2006). Such characters are by definition unreliable for cryptic species and genetic methods may then be necessary to confirm species identification. For example, Bombus hortorum and Bombus ruderatus have long been recognised as distinct species (see Williams and Hernandez 2000; Cameron et al. 2007). Their identification based on morphology alone is difficult even for experienced field biologists whereas they can be easily separated by the mitochondrial cytochrome oxidase II and cytochrome b genes (Ellis et al. 2005, 2006).

The subgenus Bombus sensu stricto (synonym Terrestribombus) is a widespread and commercially exploited holarctic taxon of bumble bee. Yet despite their large size, abundance and conspicuous colouration, the number of species across its range is still unresolved. Morphological classification of species has been complicated primarily by subtle intraspecific variation (e.g. historically there have been 186 synonyms for B. lucorum alone; Williams 1998). This has obvious implications for understanding the biology and the conservation status of the constituent species of Bombus s.str.

In Europe, five species of Bombus s.str. are known: B. cryptarum, B. lucorum, B. magnus, B. terrestris and B. sporadicus. The taxonomic status of B. lucorum, B. terrestris and B. sporadicus is widely accepted (Williams 1998). Bombus sporadicus is morphologically distinct from the other four European members of the subgenus, it is genetically distant to them (Cameron et al. 2007) and it is restricted in its European distribution to Fennoscandia (Løken 1973). Difficulties arise over the identification of the other four species that are far more widely distributed across Europe. In the British Isles, queens of B. terrestris are morphologically distinct from those of B. cryptarum, B. lucorum and B. magnus (e.g. Benton 2006, in which the latter three are synonymised to B. lucorum); the latter three species form the ‘lucorum complex’ of cryptic species.

Queens of B. cryptarum, B. lucorum and B. magnus have been differentiated by subtle morphological characters (Rasmont 1984; Pamilo et al. 1997; Bertsch et al. 2004 and refs. therein). Yet Williams (2000) considered these three taxa to represent one species, B. lucorum, as there was a morphological continuum of queens between species (see Westrich 1989, p. 582, for an historical account of the taxonomy of the lucorum complex). Recently, however, B. cryptarum and B. magnus have been demonstrated to differ subtly in queen morphology, male labial gland secretions and mitochondrial DNA from each other, albeit based on limited sampling from only two geographic locations (n = 2 sequences per species from each of two localities) and from B. lucorum (Bertsch 1997a; Bertsch et al. 2004, 2005). In the field, morphological characters do not allow ready identification of queens whilst workers and males are morphologically indistinguishable (Alford 1975; Rasmont 1984; Rasmont et al. 1986). Identification in the field is further complicated by the fact that workers and males of many subspecies of B. terrestris are also morphologically indistinguishable from those of the lucorum complex, a problem compounded by the high levels of intraspecific variation found within each species across its geographic range (Alford 1975; Williams 1991, 1998).

Accordingly, ecological inference in the lucorum complex is problematic when species delineation is unclear. In Britain, B. magnus is considered a species of upland moor and heathland sites to the North and West (Alford 1975; IBRA 1980; Prŷs-Jones and Corbet 1991) whereas in Germany it is considered more of a lowland species (von Hagen 2003). Bombus lucorum has been considered widespread and abundant, even in urban parks and gardens across Europe (Prŷs-Jones and Corbet 1991; von Hagen 2003; Benton 2006). Little has been reported on the ecology of B. cryptarum, though it has been considered to be an early spring species whose phenology precedes that of B. lucorum (Bertsch 1997b) and B. magnus (Bertsch et al. 2004). In Germany it is thought to be widely distributed and even to reach altitudes of 2700 m (von Hagen 2003). Thus, although the lucorum complex may be segregating along an altitudinal gradient, the absence of a method to clearly identify the sympatric taxa involved precludes any investigation of the constituent species’ autecology.

Consequently, previous studies on the conservation threat to these taxa may have overlooked the presence of cryptic species diversity. Based on 11 recent Regional Red Data Lists of bees across Europe, compiled according to criteria of the IUCN (2001, 2003, 2006) or analogous methods, B. lucorum is believed not to be of conservation concern. As yet, B. cryptarum remains to be cited by any but the Irish red data list (Fitzpatrick et al. 2006), and only three out of the 11 lists include B. magnus, where it is considered endangered in the Netherlands and ‘data deficient’ in Germany and Ireland. Given the increased trade in members of the subgenus Bombus as commercial pollinators (Velthuis and van Doorn 2006) and the concern over declines in bumblebees across Europe (e.g. Carvell et al. 2006; Fitzpatrick et al. 2007; Kosior et al. 2007), the need for a rigorous taxonomy of this ecologically and economically important group is becoming increasingly apparent.

To reduce taxonomic uncertainty and to facilitate the accurate determination of the conservation status and ecology of the constituent taxa within the lucorum complex, we produced a well supported phylogeny of European members of the subgenus Bombus, quantifying the degree of intra- and interspecific variation and providing strong support for the species status of the constituent taxa. We then addressed the problem of identifying cryptic species by a relatively quick and economic restriction enzyme based mtDNA marker system that reliably differentiates B. cryptarum, B. lucorum, B. magnus and B. terrestris across their European ranges. Finally, using the newly developed mtDNA marker system, we assessed the distribution and abundance of the constituent sympatric taxa of the lucorum species complex along an altitudinal gradient in Ireland.

Materials and methods

Field collection of specimens

In total 391 individuals, comprising queens and workers, of the lucorum complex of bumble bees were collected in both rural and urban environments across Ireland in 2005 and 2006 from flowers (Table 1); they have a distinctive black and yellow ‘thorax’ (mesosoma) and black, yellow and white ‘abdomen’ (metasoma or gaster) (see e.g. Benton 2006). Sampling sites included an altitudinal gradient up the Wicklows, an upland range of hills close to Dublin on the east coast of Ireland that rises to 926 m, as well as a mix of rural and urban sites. An additional 12 queens from across Europe, morphologically confirmed to species, were kindly provided by Andreas Bertsch. Individuals were either frozen or stored in 99% ethanol at 4°C prior to DNA extraction.
Table 1

Incidence of the three cryptic Bombus species in Ireland, as determined by RFLP analysis

Locality code

Dates

Location

Latitude

Longitude

Altitude a.s.l. (m)

n

B. cryptarum

B. lucorum

B. magnus

BF

10–27/04/2006

Belfast City, Co. Antrim

54°34′13 N

05°55′09″ W

5

23 Qa

5

12

0

21/06–03/07/2006

17 Wa

5

18

0

BB

05/04/2005

Benbulben, Co. Sligo

54°18′49′′ N

08°23′13″ W

350

56 W

10

28

18

SG

06/06/2005

Slieve Gullion, Co. Down

54°06′47′′ N

06°24′55′′ W

300

57 W

15

24

18

DB

03–04/2006

Dublin City, Co. Dublin

53°20′22′′ N

06°13′38′′ W

5

40 Q

11

29

0

GM

03–04/2006

Glenasmole, Co. Dublin

53°13′54′′ N

06°20′52′′ W

200

25 Q

3

13

9

GC

27/04/2006

Glencree, Co. Wicklow

53°11′41′′ N

06°18′22′′ W

350

20 Q

9

4

7

KP

18/05/2006

Kippure, Co. Wicklow

53°10′47′′ N

06°18′18′′ W

550

20 Q

6

3

11

PC

13/04/2005 and 27/04/2006

Powerscourt, Co. Wicklow

53°09′59′′ N

06°14′55′′ W

200

9 Q and 10 Q

2 3

2 4

5 3

CL

26–27/04/2006

Clara, Co. Wicklow

52°58′07″ N

06°15′59′′ W

150

33 Q

2

14

17

KL

11/05/2005

Killarney, Co. Kerry

52°03′48″ N

09°29′55′′ W

50

35 Q

1

22

12

CK

19–23/04/2006

Cork City, Co. Cork

51°53′54″ N

08°25′29′′ W

5

46 Q

0

44

2

     

Total

130 W, 261 Q

72

217

102

aCaste: Q—queen; W—worker

DNA extraction and amplification

DNA was extracted from a single leg using 10% Chelex (Walsh et al. 1991) or from half a thorax using a standard high salt protocol (Paxton et al. 1996). To avoid destructive sampling, extractions based on the terminal tarsus of either left or right midlegs were also used. Partial mitochondrial CO1 (1064 bp) was amplified by PCR using primers originally developed for Apis mellifera (Tanaka et al. 2001: forward 5′-ATAATTTTTTTTATAGTTATA-3′ and reverse 5′-GATATTAATCCTAAAAAATGTTGAGG-3′). Each 15 μl reaction contained: 1x Promega PCR buffer (containing 1.5 mM MgCl2), 60 μM dNTPs, 0.4 μM each primer and 0.6 U of Taq polymerase. PCRs were performed in an MJ PTC-100 Peltier thermal cycler under the following conditions: one denaturing cycle for 1 min at 93°C; 30 cycles of 45 s at 93°C, 1 min at 45°C and 3 min at 60°C, followed by a final extension cycle of 4 min at 60°C. PCR products were cycle sequenced using Big Dye termination chemistry (Applied Biosystems) with the same primers as for PCR and resolved on an autosequencer (Genetic Analyser 3130XL, Applied Biosystems).

Phylogenetic analyses

Published CO1 sequences of B. cryptarum, B. lucorum, B. magnus and B. terrestris (n = 25 individuals, see Table 2) as well as our own sequences from morphologically identified specimens from across Central and Northern Europe (n = 12 non-Irish individuals, see Table 2) were aligned using Clustal W (Thompson et al. 1994). Phylogenetic trees were generated using both mega 3.1 (Kumar et al. 2004) and mrbayes 3.1.2 (Ronquist and Huelsenbeck 2003) and viewed using treeview 1.6.6 (Page 1996). The CO1 sequences of B. sporadicus (GenBank Accession Nos. AF279500 and AY181163) were used as the outgroup. In mega, the maximum parsimony method was applied using the Close-Neighbour-Interchange (CNI) branch-swapping heuristic search with random addition of initial trees, replicated 10 times for each run and assessed by bootstrap for 1,000 replicates. The maximum likelihood trees generated using the Bayesian MCMC (Markov Chain Monte Carlo method) analysis of mrbayes was based on the general time reversible model (GTR) of base substitution, gamma distribution, 100,000 generations to achieve stationarity, sampling every 10 generations and a ‘burn-in’ of 2,500.
Table 2

RFLP haplotypes of mitochondrial CO1 of B. lucorum, B. cryptarum, B. magnus and B. terrestris

Species

GenBank Acc. No.

Baseb

2150

2249

2456

2524

2537

2617

2630

2816

2966

2975

Haplotype code

Restriction enzymec

E

H

E

H

H

H

H

H

H

H

Locality

B. lucorum

AY181117

Kaprun, Austria

x

         

A

AY181118

Färnigen, Switzerland

x

         

A

AY181118

Bivio, Switzerland

x

         

A

AY181118

Sattelegg, Switzerland

x

         

A

AY181119

Tastrup, Denmark

x

         

A

AY181119

Særløse Overdrev, Denmark

x

         

A

AY181120

Al, Buskerud, Norway

x

         

A

AY181120

Bröstrud. Norway

x

         

A

AY181120

Skute, Norway

x

         

A

AY181120

Hven, Sweden

x

         

A

AY530009

Menz, Brandenburg, Germany

x

         

A

AY530010

Menz, Brandenburg, Germany

x

         

A

EF362739a

Crail, Fife, Scotland

x

         

A

EF362739a

Benbulben, Sligo, Ireland

x

         

A

B. cryptarum

AY181123d

Sölk Pass, Austria

x

 

x

 

x

     

B

AY181124d

Julierpass, Switzerland

x

 

x

 

x

     

B

AY181124d

Bivio, Switzerland

x

 

x

 

x

     

B

AY181124d

Gadmen, Switzerland

x

 

x

 

x

     

B

AY530012

Menz, Brandenburg, Germany

x

 

x

 

x

     

B

AY530013

Menz, Brandenburg, Germany

x

 

x

 

x

     

B

EF362728a

Leerstetten, Germany

x

 

x

 

x

     

B

EF362728a

Marburg, Germany

x

 

x

 

x

     

B

EF362729a

St. Petersburg, Russia

x

 

x

 

x

 

x

   

C

AY530011

Duncansby Head, Scotland

x

 

x

 

x

 

x

   

C

EF362726a

Orkney Is., Scotland

x

 

x

 

x

 

x

   

C

EF362726a

Slieve Gullion, Armagh, Northern Ireland

x

 

x

 

x

 

x

   

C

EF362727a

Benbulben, Sligo, Ireland

x

  

x

 

x

    

D

B. magnus

AY530015

Menz, Brandenburg, Germany

x

 

x

   

x

   

E

EF362737a

Teupitz, Brandenberg, Germany

x

 

x

   

x

   

E

EF362733a

Teupitz, Brandenberg, Germany

x

 

x

   

x

 

x

 

F

EF362733a

Milde, Bergen, Norway

x

 

x

   

x

  

x

G

AY530014

Duncansby Head, Scotland

x

 

x

   

x

   

E

EF362736a

Brecon Beacons, Wales

x

 

x

   

x

   

E

EF362738a

Exmoor, England

x

 

x

   

x

   

E

EF362735a

Benbulben, Sligo, Ireland

x

 

x

   

x

   

E

B. terrestris

AY181169

Denmark

x

x

     

x

  

H

AY181170

Denmark

x

x

     

x

 

x

I

AY181171

Edinburgh, Scotland,

x

x

x

    

x

  

J

EF362743a

Marburg, Germany

x

x

       

x

K

EF362744a

St. Petersburg, Russia

x

x

x

      

x

L

EF362742a

Cork City, Cork, Ireland

x

x

     

x

 

x

I

EF362741a

Belfast, Antrim, Northern Ireland

x

x

     

x

 

x

I

New GenBank Accession numbers are indicateda and their RFLP haplotype codes have been confirmed by agarose gel electrophoresis (Fig. 3). Other haplotype codes have been deduced in silico from GenBank Accession sequences

bPosition based on the published CO1 sequence of Apis mellifera ligustica (Crozier and Crozier 1993)

cE—EcoN1; H—Hinf1

dErroneously given as B. magnus in GenBank (see Bertsch et al. 2005 for explanation based on the distribution of B. magnus)

To place Irish specimens within the phylogenies generated above and to increase sample sizes for genetic diversity analyses, we added 20 Irish specimens (GenBank Accession Nos. EF362725–EF362744 inclusive), unambiguously identified using the RFLP protocol described below (n = 6 B. cryptarum; n = 5 B. lucorum; n = 7 B. magnus; n = 2 B. terrestris) to the sequence dataset, giving a total of 59 sequences from the five species (Table 2). Irish specimens were either identical to previously published sequences or varied by only 1–6 SNPs from them, allowing clear species identification. Additional phylogenetic trees were then generated with the entire dataset of 59 sequences using the approaches and parameter values detailed above.

The absolute number of substitutions within and among species and the net Tamura–Nei genetic distance was calculated (as interspecific distance minus intraspecific variation) with standard errors (1,000 bootstrap replicates) using mega 3.1 (Kumar et al. 2004). The Tamura–Nei model of base substitution was chosen as it accounts for the strong A + T bias found in CO1 genes of Hymenoptera (Crozier and Crozier 1993; Tamura and Nei 1993; Simon et al. 1994).

RFLP protocol for species identification

The 16 unique CO1 sequences of B. cryptarum, B. lucorum, B. magnus and B. terrestris (see Table 2 for GenBank Accession Nos.) were screened to identify potential discriminatory restriction sites in our 1064 bp amplicon using seqbuilder (lasergene 6.1; DNAStar Inc., Madison, Wisconsin). Sequences excluded those derived from Irish specimens but were otherwise derived from across the Central and North European ranges of all species (Fig. 1). Both EcoNI and HinfI were selected as a single restriction enzyme could not differentiate the four species. PCR products were digested with both restriction enzymes (New England Biolabs), each reaction containing 15 l PCR product, 1 U of both enzymes and 1× NEBuffer number 2. Reaction volumes were made up to 20 μl with H20, gently mixed and centrifuged, then incubated at 37°C for 4 h. Fragments were resolved by electrophoresis in 2% agarose gels for 2.5 h at 100 V, stained with ethidium bromide and photographed. Both previously sequenced individuals (DNA extracts) and a 100 bp ladder (Invitrogen) were used to confirm fragment sizes.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9394-z/MediaObjects/10592_2007_9394_Fig1_HTML.gif
Fig. 1

Sampling localities for B. cryptarum, B. lucorum, B. magnus and B. terrestris. Black dots: own sequences and RFLP data; white dots: sequence data from GenBank

An Irish individual with the one novel RFLP (Haplotype Code D, see Table 2) was sequenced to determine its identity. Five to seven other Irish specimens were sequenced per species of the lucorum complex to ensure that RFLPs were consistently associated with the same or were very similar (within six SNPs) to CO1 sequences of morphologically identified specimens. Their data were included in the analyses of genetic diversity (see above).

Distributional analyses

Based on results of the RFLP analyses, the proportions of B. cryptarum, B. lucorum and B. magnus were calculated for each location sampled in Ireland. We then undertook two distributional analyses. In the first, we compared differences in the numbers of each species at rural sites across our Wicklows altitudinal gradient (at or below versus above 200 m) by a G goodness of fit test. We then compared differences in the numbers of each species at urban versus low-altitude (≤200 m) rural sites with a G-test.

Specimens are housed at the University of Dublin Trinity College (with MJFB) and DNA extracts at Queen’s University Belfast (with RJP). Throughout, means are presented ± SE.

Results

How many cryptic species within the lucorum complex?

Alignment of the 59 partial CO1 sequences was simple as divergence was slight (maximum 8.3% between the most divergent sequences), there were no indels, and all sequences contained an open reading frame. Based on sequences trimmed to 700 bp from these 59 specimens distributed across Central and North Europe (Fig. 1), there were 85 parsimony informative sites. The 20 Irish sequences were either identical to or very similar to sequences of B. cryptarum, B. lucorum, B. magnus or B. terrestris. All 20 contained an open reading frame and could be aligned with the other 39 sequences (no indels). Phylogenetic analysis of the dataset revealed the Irish specimens to be clearly embedded within one of four taxa (Fig. 2). The topology of both the cladogram (Fig. 2a) and phylogram (Fig. 2b) support B. cryptarum and B. magnus as distinct monophyletic taxa, separate from B. lucorum, with strong support for the species level nodes using both approaches. Within the 700 bp of aligned sequence there were 3–22 diagnostic sites per species (i.e. 3–22 sites with a DNA base unique to only one of the five species, see Table 3). The phylogenetic trees, generated using samples of four taxa distributed across their European ranges where they exist in sympatry, and the presence of fixed diagnostic sites between the taxa (Table 3), provide strong support for the existence of B. cryptarum, B. lucorum, B. magnus and B. terrestris as species that are discrete genotypic clusters in sympatry (Mallet 1995, 2007). The topology of the tree suggests that within B. cryptarum, with an unresolved Irish-Scottish clade and a Continental European clade, there may be some geographic clustering (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9394-z/MediaObjects/10592_2007_9394_Fig2_HTML.gif
Fig. 2

(a) Cladogram and (b) phylogram of B. cryptarum, B. lucorum, B. magnus and B. terrestris mitochondrial CO1 based on 700 bp of aligned sequences, using B. sporadicus as outgroup. The 59 sequences includes 20 Irish specimens. (A) Maximum parsimony tree generated by using CNI branch-swapping heuristic search with random addition of initial trees, replicated 10 times for each run (for parsimony informative sites: CI = 0.759, RI = 0.961, RCI = 0.730). Numbers above branches indicate bootstrap support based on 1,000 bootstraps. (B) Maximum likelihood tree obtain by Bayesian MCMC analysis with the general time reversible model, gamma distribution and 100,000 generations. Numbers at the nodes indicate posterior probabilities of each split. Scale bar refers to the expected number of substitutions per site. See Table 2 for collection locality. * Indicates an Irish specimen

Table 3

Diagnostic sites in the 700 bp partial sequence of CO1 for all the individuals listed in Table 2

 

Nucleotide positiona

 

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

4

4

4

4

4

 

8

8

0

1

1

2

2

2

3

3

3

9

1

2

4

6

6

9

2

3

5

8

9

 

6

9

1

3

6

0

5

8

1

4

7

7

2

7

5

0

9

4

0

8

9

6

5

B. sporadicusa

A

T

C

A

T

T

A

T

A

C

T

C

T

C

C

A

T

T

T

A

T

T

T

B. cryptarum

.

A

T

T

A

A

.

A

.

T

.

T

.

A

T

C

A

A

A

.

.

C

A

B. lucorum

C

A

T

C

A

A

.

A

.

T

.

T

C

A

T

T

A

A

A

T

.

C

A

B. magnus

.

A

T

T

A

A

T

A

.

T

.

.

.

A

T

T

A

A

A

.

C

C

A

B. terrestris

.

.

T

T

A

A

.

A

C

T

A

T

.

A

T

T

A

.

.

.

.

C

A

 

5

5

5

5

5

5

5

6

6

6

6

6

6

7

7

7

7

7

7

8

8

8

8

 

1

2

2

3

6

6

9

1

3

5

6

8

9

0

1

3

3

4

5

1

3

4

5

 

9

3

8

5

6

7

4

2

6

1

3

7

0

5

1

2

6

1

3

9

8

0

5

B. sporadicusa

T

T

T

A

T

T

C

T

T

T

C

A

A

A

A

A

A

T

A

A

T

A

T

B. cryptarum

C

C

.

.

C

C

T

C

.

A

A

T

T

T

T

T

.

.

T

.

C

T

.

B. lucorum

.

.

.

G

C

C

T

.

C

.

A

T

T

T

T

.

.

.

T

T

C

T

.

B. magnus

.

.

.

.

C

C

A

C

.

.

A

T

T

T

T

T

C

.

T

T

C

T

C

B. terrestris

.

.

C

.

C

C

T

.

.

A

T

.

T

C

T

.

.

A

T

.

C

T

C

Dots (.) indicate sequence identity to the reference sequence of B. sporadicus on the first line

aNucleotide positions based on the published CO1 sequence of B. sporadicus (GenBank Accession Number: AY181162; Pedersen 2003)

The intraspecific genetic divergence among species (mean 0.35%, Table 4) was an order of magnitude lower than net interspecific variation (mean 5.86%, Table 5). Within this subgenus, there was a clear ‘barcoding gap’ of sequence divergence between species. The net genetic distances between the three members of the lucorum complex, namely B. lucorum, B. magnus and B. cryptarum, were approximately half that of the corresponding distances to the sister species B. terrestris and considerably lower than to B. sporadicus, the other European member of the subgenus Bombus (Table 5).
Table 4

Intraspecific genetic variation for the five European species of the subgenus Bombus s.str. based on 700 bp sequences of partial CO1

 

n

Distance

n substitutions

B. cryptarum

17

0.004 ± 0.001 (0.000–0.018)

2.9 ± 0.8 (0–12)

B. lucorum

19

0.002 ± 0.001 (0.000–0.007)

1.4 ± 0.6 (0–6)

B. magnus

14

0.004 ± 0.001 (0.000–0.010)

5.3 ± 0.9 (0–9)

B. terrestris

7

0.013 ± 0.003 (0.000–0.023)

8.6 ± 1.9 (0–15)

B. sporadicus

2

0.006

4.0

Intraspecific genetic distance ± SE (Tamura and Nei 1993) and mean number of substitutions ± SE among haplotypes, with ranges in parenthesis

Table 5

Interspecific genetic divergence between the five European species of the subgenus Bombus s.str. based on 700 bp sequences of partial CO1 (see Table 2 for sample sizes)

 

B. cryptarum

B. lucorum

B. magnus

B. terrestris

B. sporadicus

B. cryptarum

0.036 ± 0.007 (0.035–0.049)

0.023 ± 0.007 (0.026–0.035)

0.055 ± 0.010 (0.038–0.062)

0.080 ± 0.012 (0.084–0.091)

B. lucorum

23.0 ± 4.3 (23–31)

0.035 ± 0.007 (0.034–0.051)

0.060 ± 0.010 (0.047–0.086)

0.083 ± 0.012 (0.083–0.097)

B. magnus

16.2 ± 3.9 (17–23)

21.8 ± 4.3 (23–31)

0.053 ± 0.010 (0.038–0.072)

0.083 ± 0.012 (0.078–0.094)

B. terrestris

32.5 ± 5.1 (25–42)

35.9 ± 5.3 (30–50)

30.8 ± 5.1 (25–43)

0.078 ± 0.013 (0.072–0.096)

B. sporadicus

48.3 ± 6.4 (51–54)

49.9 ± 6.3 (51–58)

48.5 ± 6.3 (48–56)

44.4 ± 6.0 (44–55)

Top diagonal: net mean interspecific genetic distance ± SE (Tamura and Nei 1993); bottom diagonal: net mean number of substitutions ± SE; and gross interspecific ranges in parenthesis

Analysis of the number of substitutions within (Table 4) and between (Table 5) species obviously revealed the same pattern as genetic distance measures. Firstly, intraspecific variation (Table 4) was far lower than interspecific differences (Table 5). Secondly, there were over 1.5 times the number of substitutions between B. terrestris and the lucorum complex (mean 33.1 substitutions), relative to the number of substitutions among species within the lucorum complex (mean 20.3 substitutions, Table 5). Based on both genetic distance and number of substitutions, B. cryptarum and B. magnus were the closest species pair, yet still the genetic distance between these species was over four times greater than the divergence among haplotypes within either of these two species (cf. Tables 4, 5). Although B. terrestris was not a focal species in this study, it interestingly appears to have greater intraspecific haplotype diversity compared to the other taxa (Table 4), despite a smaller sample size.

RFLP haplotype diversity and distribution

Restriction fragments summed to the total amplicon length, suggesting that we had not amplified shorter amplicons that might have been indicative of nuclear pseudogenes (numts). RFLPs unequivocally discriminated B. cryptarum, B. lucorum, B. magnus and B. terrestris in silico for all published sequences and by agarose gel electrophoresis for all non-Irish individuals morphologically identified and sequenced by ourselves (Table 2). Of the 391 Irish haplotypes, 389 were identical to those of morphologically identified specimens, allowing unambiguous species designation. Only one unique haplotype, found in two individuals, was detected in the Irish dataset, (Haplotype Code D, see Table 2). It differed by only three SNPs from other B. cryptarum sequences and possessed the unique species-diagnostic bases, and it was clearly embedded within the B. cryptarum clade (Fig. 2).

Using our technique, B. lucorum appears to have a single haplotype across the eight countries sampled including Ireland (Code A), its banding pattern being easily distinguished from that of the other taxa investigated, as confirmed by both sequencing and RFLPs (Table 2, Fig. 3). The distribution of B. cryptarum haplotypes mirrors that of the tree topologies (Fig. 2) in that the haplotype found exclusively in mainland Europe (Code B) differs from the British and Irish haplotypes (Codes C and D, Table 2). Within Ireland, the B. cryptarum haplotype Code C was dominant and only two individuals had the haplotype Code D. Bombus magnus was represented by one haplotype (Code E) in our Irish samples whereas elsewhere in Europe three haplotypes were found (Codes E–G). Bombus terrestris exhibited five haplotypes (Codes H–L inclusive) among a sample of only seven individuals.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9394-z/MediaObjects/10592_2007_9394_Fig3_HTML.jpg
Fig. 3

Ethidium bromide stained agarose gel (2%) with examples of RFLP haplotypes (haplotype codes given as letters at the bottom of each lane, see Table 2) found in each taxon: B. lucorum, lanes 2 (Ireland), 3 (Germany) and 4 (Germany); B. cryptarum, lanes 5 (Ireland), 6 (Ireland), 7 (Scotland), 8 (Germany), 9 (Germany), 10 (Russia) and 11 (Russia); B. magnus, lanes 12 (Ireland), 13 (England), 14 (Germany), 15 (Germany) and 16 (Germany); B. terrestris, lanes 17 (Ireland), 18 (Canary Isles), 19 (Germany) and 20 (Russia); lane 21, uncut sample; lanes 1 and 22, 100 bp ladder

Distribution of the cryptic lucorum complex of species within Ireland

All three species of the lucorum complex were widespread and occurred in 8 out of 11 Irish sites (Fig. 3). Bombus cryptarum was absent from only one site (CK) whereas B. magnus was absent from two of the urban sites, DB and BF, and was found at very low frequency in the third urban site, CK (Fig. 4). B. lucorum was found at all sites and comprised 55.5% of Irish samples, followed by B. magnus (26.1%) and B. cryptarum (18.4%; see Table 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9394-z/MediaObjects/10592_2007_9394_Fig4_HTML.gif
Fig. 4

Map of Ireland showing the sampling localities, sample size, height (in m) above sea level and relative proportions of B. cryptarum (grey), B. lucorum (black) and B. magnus (white). See Table 1 for locality codes and details

Along the Wicklows altitudinal gradient (150–550 m) of entirely rural sites there was significant heterogeneity in species composition (G2 = 12.460, P = 0.002); B. cryptarum was relatively more frequent at sites above 200 m, B. lucorum was relatively less frequent above 200 m and B. magnus showed no difference in its relative altitudinal distribution (Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9394-z/MediaObjects/10592_2007_9394_Fig5_HTML.gif
Fig. 5

Bombus cryptarum, B. lucorum and B. magnus along an altitudinal gradient in the Wicklows from site CL to KP (see Table 1 for locality codes)

Comparison of urban (n = 3) and low altitude (≤200 m) rural (n = 4) sites also revealed significant heterogeneity in species composition (G2 = 59.772, P < 0.0001); B. cryptarum and particularly B, magnus were relatively more abundant in rural sites rather than urban sites. Bombus lucorum was relatively more frequent at urban sites compared to the other two species. Yet it was in fact ubiquitously distributed in both urban and rural sites (Fig. 4).

Discussion

We have provided strong support for the view that the lucorum complex from Central and Northern Europe comprises three monophyletic bumble bee taxa (B. cryptarum, B. lucorum and B. magnus) and that the broader subgenus Bombus s.str. includes a fourth species across the same range (B. terrestris), with an additional fifth species in Fennoscandia (B. sporadicus). Our PCR-RFLP method allows accurate (100% correspondence between RFLP patterns and CO1 sequences) rapid (within 10 h of DNA extraction) and relatively cheap discrimination (per sample cost after PCR is ≈25 times less expensive than sequencing) of the three species of the lucorum complex and B. terrestris in Europe, where morphological identification of queens, workers or males is often problematic or impossible, even for experienced taxonomists. It is therefore an indispensible tool in assessment of the conservation status of constituent species of Bombus s.str. in Europe.

In Russia and Japan, the technique can still discriminate between the lucorum complex, B. terrestris and Bombus patagiatus, the latter being another member of Bombus s.str. But two cryptic Asiatic sister species to B. cryptarum, namely Bombus albocinctus and Bombus florilegus, theoretically produce RFLP bands identical to those of known B. magnus haplotypes (TM, unpublished data). However, the species status of these Asiatic taxa is as yet uncertain. The situation is complicated by the possibility of many more taxa of the lucorum complex in Asia (Williams 1991, 1998).

Our support for the view that the lucorum complex in Europe comprises three species rests on both the presence of fixed diagnostic nucleotide differences between the taxa (Table 3) and the low level of intraspecific sequence variation in CO1 (≤1.3%) despite the higher level of interspecific divergence (2.3–8.3%) in specimens collected from across much of their sympatric European range. CO1 has become widely used in insect systematics (Caterino et al. 2000) and more recently in barcoding (Hebert et al. 2003a, b). Hebert et al.’s (2004) study of 10 cryptic butterfly species found interspecific CO1 divergence to vary considerably, from 0.3% to 6.6%. More recently, Meyer and Paulay (2005) have warned of difficulties in the use of CO1-based barcodes to differentiate among taxa when prior sampling of taxa is incomplete. Furthermore, many authors have criticised the use of distance-based methods of species delimitation using DNA sequences due to the lack of integration with diagnostic characters used in classical taxonomy (DeSalle et al. 2005, DeSalle 2006) and the difficulty in standardising the percent sequence divergence to diagnose species (Cognato 2006). We have nevertheless found CO1 sequences to delineate well among closely related cryptic species in a formerly poorly characterised bumble bee complex, with a clear ‘barcoding gap’ between species, as well as 46 sites with which species within this taxon may be diagnosed (and 3–22 unique sites per species). We suggest that the integration of character-based and distance-based barcoding methods may be a valuable approach to the identification of bumble bees, a widely distributed, abundant and ecologically important group of insects.

That we detected reciprocal monophyly in the CO1 sequences among the three taxa of the lucorum complex when in sympatry further supports the view that they represent good genetic species (Mallet 1995, 2007). The character-based parsimony and likelihood phylogenies based on the sequences were, moreover, very well supported. Reciprocal monophyly also indicates that any shared ancestral polymorphisms have been lost, either as a consequence of stochastic lineage pruning or selective sweeps (Bazin et al. 2006). It will now be important to determine the extent to which nuclear gene pools of the three taxa of the lucorum complex are isolated (e.g. Monaghan et al. 2005), as discrepancies between mitochondrial and nuclear phylogenies are not uncommon (Funk and Omland 2003).

A recent comprehensive phylogeny of bumble bees (Cameron et al. 2007), based on one mitochondrial (16S rRNA) and four nuclear genes, revealed monophyly of the subgenus Bombus and, more importantly, found similar mitochondrial DNA divergence between B. cryptarum and B. lucorum (2.6%) as we did using CO1 gene sequences (3.6%). The interspecific CO1 divergence among the lucorum complex revealed by Bertsch et al. (2005) was also similar to that we report here. These sequence data of numerous individuals sampled across their European ranges, combined with the species specific odour blends of male labial gland secretions of the lucorum complex (Bertsch 1997a; Bertsch et al. 2004, 2005), lend further weight to the view (Rasmont 1984; Rasmont et al. 1986) that B. cryptarum, B. lucorum and B. magnus are distinct species.

Bombus cryptarum was earlier believed to be absent from both Britain and Ireland (Alford 1975; Rasmont 1984). Bertsch et al. (2005) recently identified B. cryptarum from one county in Scotland. Here we have shown that the species is well distributed throughout Ireland as well as other Central and Northwest European countries. Bombus lucorum and B. magnus are similarly well distributed across Central and Northwest Europe, including Ireland.

Our data on the relative proportion of each species of the lucorum complex in Ireland is a crude assessment of the incidence of those species, as proportions will vary both temporally and spatially due to species differences in emergence times (e.g. Bertsch et al. 2004) and levels of parasitism. Nevertheless, we can state with some confidence that, within Ireland, B. lucorum is the most abundant of the lucorum complex. Using morphological identification, Banaszak and Rasmont (1994) similarly considered B. lucorum to be the most common of the lucorum complex in Poland, whilst using allozyme based identification Pamilo et al. (1997) also considered it the most abundant taxon of the complex in Finland.

Assertions that B. magnus in Britain is an upland species, preferring bog or heathland habitats (Alford 1975; Prŷs-Jones and Corbet 1991), were based on morphological characterisation of individuals. Our more accurate species designation allows us to state that, in Ireland, B. magnus is not associated with uplands per se, though it is also not associated with urban areas. The distribution of B. magnus in Germany, namely: lowland non-urban (von Hagen 2003), follows that in Ireland. The distribution of B. magnus in Britain may need to be re-appraised using our RFLP identification protocol. Of the lucorum complex in Ireland, we found B. cryptarum to be more associated with uplands.

The difficulty in separating among the lucorum complex, particularly in identifying B. cryptarum and B. magnus in the field even with freshly caught queens (e.g. in Scotland; Bertsch et al. 2005), means that potentially these cryptic species may have been overlooked in large scale population genetic studies of Bombus s.str. This may be especially pronounced in studies where the sampling design has been heavily biased towards sampling workers, rather than queens, for which there is a lack of consistent morphological characters to discriminate B. cryptarum, B. magnus, B. lucorum and even B. terrestris in many parts of Europe. Sampling populations with a mixture of species will result in greater intra-population genetic variation, potentially resulting in a spurious lack of inter-population differentiation. Consequently, the conservation status of the constituent species of the lucorum complex needs to be reappraised.

Both B. lucorum and B. terrestris are considered to be among the most common and widespread bumble bee species found in Europe (e.g. von Hagen 2003; Benton 2006). Currently B. terrestris is exploited for commercial pollination both within (Europe) and outside its natural range (Canada, Chile, China, Japan, New Zealand, South Korea, Tasmania and the U.S.A.; see Velthuis and van Doorn 2006). In Europe, North Africa and East Asia, 900,000 colonies of B. terrestris and 4,000 colonies of B. lucorum were produced for commercial pollination in 2004 alone (Velthuis and van Doorn 2006). In the laboratory, commercially imported European B. terrestris can successfully produce hybrids with native Japanese B. hypocrita sapporoensis (Mitsuhata and Ono 1996). Different subspecies of B. terrestris vary in a number of behavioural traits and frequently non-native subspecies of B. terrestris outcompete native subspecies (Chittka et al. 2004; Ings et al. 2005, 2006), as well as transmit pathogens to both conspecifics and congeners (Goka et al. 2001; Colla et al. 2006). These concerns highlight the necessity of correctly identifying native taxa. As the commercial production of Bombus is set to increase (Velthuis and van Doorn 2006), it is essential that any taxonomic confusion be resolved to aid investigation of the impact of introduced bumble bees on native fauna through potential genetic homogenisation of indigenous bumble bee species and pathogen transmission from commercially introduced bumble bees. Our RFLP approach to discrimination among the lucorum complex and B. terrestris, when combined with DNA extraction from tarsal samples (Holehouse et al. 2003) that removes the necessity for selectively or destructively sampling adults to confirm taxonomic identity, provides a powerful tool to reduce taxonomic uncertainty so as to analyse the effects of bumble bee commercialisation and the conservation status of important native fauna.

Acknowledgements

We thank friends and colleagues who helped to collect bumble bees across Ireland: D. Cookson, D. Dominoni, M. Kelly and S. Roos; and Andreas Bertsch for provision of additional samples, discussion and encouragement to engage with the lucorum complex. We also thank Andreas Bertsch, Jim Provan, Alfried Vogler, Paul Williams and an anonymous reviewer for many helpful comments on the manuscript. This work was supported by a grant from the Higher Education Authority of Ireland as part of its North-South Research Programme for Peace and Reconciliation.

Copyright information

© Springer Science+Business Media B.V. 2007